First complete pterosaur from the Afro-Arabian continent: insight into pterodactyloid diversity

Despite being known from every continent, the geological record of pterosaurs, the first group of vertebrates to develop powered flight, is very uneven, with only a few deposits accounting for the vast majority of specimens and almost half of the taxonomic diversity. Among the regions that stand out for the greatest gaps of knowledge regarding these flying reptiles, is the Afro-Arabian continent, which has yielded only a small number of very fragmentary and incomplete materials. Here we fill part of that gap and report on the most complete pterosaur recovered from this continent, more specifically from the Late Cretaceous (~95 mya) Hjoûla Lagerstätte of Lebanon. This deposit is known since the Middle Ages for the exquisitely preserved fishes and invertebrates, but not for tetrapods, which are exceedingly rare. Mimodactylus libanensis gen. et sp. nov. differs from the other Afro-Arabian pterosaur species named to date and is closely related to the Chinese species Haopterus gracilis, forming a new clade of derived toothed pterosaurs. Mimodactylidae clade nov. groups species that are related to Istiodactylidae, jointly designated as Istiodactyliformes (clade nov.). Istiodactyliforms were previously documented only in Early Cretaceous sites from Europe and Asia, with Mimodactylus libanensis the first record in Gondwana.

Diagnosis. Mimodactylid with the following autapomorphies: humerus with a rectangular deltopectoral crest; humerus smaller than half the length of the second phalanx of the wing finger (hu/ph2d4 < 0.5). This species can be further distinguished from other ornithocheiroid pterodactyloids by the following combination of Detail of the humerus. Scale-bars, a: 50 mm; b-d: 10 mm. Abbreviations. car: carpus; cdv: caudal vertebrae; cor: coracoid; cra: cranium; cs: cristospine; cv: cervical vertebrae; d: dentary; dcar: distal carpals; dpc: deltopectoral crest; dri: dorsal ribs; dv: dorsal vertebrae; fe: femur; hu: humerus; man: mandible; mcI-III: first to third metacarpals; mcIV: wing metacarpal; mtar: metatarsals; pcar: proximal carpals; ph1d4: first wing phalanx; ph2d4: second wing phalanx; ph3d4: third wing phalanx; ph4d4: fourth wing phalanx; ppu: prepubis; ptd: pteroid; ra: radius; sca: scapula; sri: sacral ribs; stp: sternal plate; sym: mandibular symphysis; tar: tarsus; te: teeth; ti: tibia; ul: ulna; the abbreviations 'l' and 'r' represents respectively left and right. characters: discrete palatal ridge; 11 and 10 cone-shaped teeth on each side of the upper and lower jaws, respectively; scapula slightly longer than coracoid; humerus much longer than femur (hu/fe ~1.3); deltopectoral crest extends for around 40% of the humerus shaft length (see Supplementary Information for further details and measurements). comparative description. The specimen is well preserved with most parts of the skeleton articulated or only slightly displaced from their anatomical position (Fig. 2a). The skull and lower jaw are exposed in ventral view, with the occipital region and the craniomandibular articulation flattened. It is a comparatively small individual, with a wingspan of ~1.32 meters, and long wings, resulting in a high aspect ratio. Based on the unfused scapula and coracoid, pelvic elements and sacral vertebrae but fused dentaries at the symphysis, and dorsal vertebrae not fused into a notarium, it is likely that it was a very young animal at the time of death, having reached an ontogenetic stage between 2 and 3 20 . The skull has a broad rostrum in dorsopalatal view (Fig. 3), but not as rounded as in Istiodactylus [21][22][23] and also differing from other istiodactyliforms 24,25 . The rostral tip is pointed, unlike the rounded terminus of Istiodactylus [21][22][23] . There are 11 and 10 cone-shaped teeth on each side of the upper and lower jaws, respectively (Fig. 3a), similar to Haopterus 26 and Linlongopterus 27 . Crowns are labiolingually compressed with a cingulum ( Fig. 3b; SI) as in Haopterus and other istiodactyliforms. This cingulum was previously reported in Istiodactylidae 28 and related taxa, but the teeth of this latter lineage is characterised by wide crowns, which have also a marked labiolingually compression [21][22][23][24][25] . The palate is concave and shows a small palatal ridge. Choanae are large and divided by the vomers. The postpalatinal fenestra has an elongated egg-like shape as in the basal istiodactyliform Hongshanopterus 25 . Ceratobranchials I of the hyoid apparatus are fork-like, thin, and elongate elements. An odontoid process is present at the tip of the lower jaw as in Istiodactylus latidens 23 , but this process could also be recognised in Haopterus 26 and Lonchodraco giganteus 29 . The dorsal vertebrae ( Fig. 2b; Supplementary Fig. S1) are exposed in ventral view and are not fused into a notarium. A total of 7 caudal vertebrae were identified ( Supplementary Fig. S2), all of which lack a duplex centrum and decrease rapidly in size posteriorly, suggesting that this species had a short tail. The cristospine of the sternum is comparatively short and deep, similar to those of Nurhachius and Istiodactylus; the anterior portion of the sternum is more rounded in lateral view than that of istiodactylids, being, in this respect, more similar to that of the Anhangueridae. The scapula is stout and shares with istiodactylids and anhanguerids a constricted shaft (Fig. 2b). However, it differs from both by being longer than the coracoid 18 . The coracoid sternal articulation is slightly concave as in Haopterus 25 and has a developed posterior expansion that is not present in istiodactylids. The humerus (Fig. 2d) has a rectangular deltopectoral crest with an unusual straight distal margin and extends approximately 40% down the humerus shaft, more than in any other ornithocheiroid except for Pteranodon and related taxa 30 . Mimodactylus has some wing elements longer relative to the humerus compared to istiodactylids, in particular the first and second phalanges. The distal portion of the last phalanx of the wing finger is curved as in most pterosaurs. The feet are relatively small, similar to istiodactylids 30 . The pteroid (Fig. 2c) is quite large (longer than the humerus). This bone is clearly articulated with the proximal syncarpal and directed towards the body. There has been a long discussion about the position of this unique pterosaur bone with the carpal elements [31][32][33] , which is clearly settled in the present specimen whose forelimb bones are perfectly articulated.

Discussion
Despite the incompleteness of the two other pterosaur specimens described from the upper Cenomanian of Lebanon, both of which lack cranial elements, those specimens can clearly be distinguished from Mimodactylus. The only complete bones of the ornithocheiroid MSNM V 38818 are the wing metacarpal and the radius + ulna, whose proportions differ from those of Mimodactylus showing that the latter has a proportionally longer forearm. Furthermore, the diameter of the radius of MSNM V 3881 is less than half that of the ulna, contrary to the condition of Mimodactylus. The second specimen, the holotype of Microtuban altivolans 16 , has a much shorter wing, a humerus with a different deltopectoral crest and a scapula that lacks the constricted shaft observed in Mimodactylus.
Another interesting feature of Mimodactylus is the dentition (Figs. 3a and SI), which differs from that of most ornithocheroids. As in Haopterus 25 and Linlongopterus 26 , the new species has cone-shaped dental crowns and they are confined to the anterior half part of the jaws. Such a configuration is present in other istiodactyliforms and cannot be established in Lonchodraco giganteus due to preservation 29 , which also present cone-shaped teeth. As in Haopterus but unlike Linlongopterus and Lonchodraco, teeth are characterised by a cingulum at the base of the crown (Supplementary Fig. S3), which is also present in the Istiodactylidae and closely related species. Mimodactylus, however, lacks the lancet-shaped teeth with marked labiolingually compressed crowns that are diagnostic of the istiodactylids [20][21][22][23][24] . The new species also lacks the sharp carinae reported in Istiodactylus 21 . The first upper tooth of Mimodactylus is comparatively small and has a sub-circular transverse section. It is followed by the largest teeth in the upper jaw, which have slight labiolingually compressed crowns with a cingulum, convex labial surfaces and thin, lingually inclined, needle-like tips. This general morphology is present in the remaining teeth, also from the lower jaw. This kind of dentition is more similar to that of the basal archaeopterodactyloids 18 Pterodactylus and Germanodactylus rhamphastinus 24 than to istiodactylids and ornithocheiroids. The sole other derived pterodactyloid with a comparable dentition is Haopterus gracilis, first regarded as an archaeopterodactyloid 26 , later as an ornithocheroid close to Istiodactylidae 28 , and even the sister taxa of Ornithocheiroidea 13,19 . Here we recovered Haopterus at the base of Laceodontia as in more recent phylogenetic analyses 34,35 , forming a clade with Mimodactylus libanensis (Fig. 4a; see Supplementary Information for further details).
The Mimodactylidae clade nov. is diagnosed by the following synapomorphies: cone-shaped teeth on each side of the upper jaws, crowns with a slight labiolingual compression, and sternal articular surface of the coracoid slightly concave. In addition, all mimodactylids have the teeth confined to the anterior half of the jaws and are widely spaced. Although several characters presented previously are recorded in mimodactylids and other istiodactyliforms, some of them are recovered as symplesiomorphies shared by other lanceodonts (e.g., character state 77(1): presence of an odontoid process in the lower jaw; character state 95(1): cone-shaped teeth) or are of unclear origin (e.g., character state 15(3): elongated egg-shaped postpalatine fenestra; character state 56 (3): quadrate inclined about 150° posteriorly to ventral margin of the skull). In addition, two synapomorphies that support the Mimodactylidae should be considered with caution due to the missing data in most of lanceodontians 30 (see Supplementary Information for further details) and the potential of ontogenetic variation in derived pterodactyloids 18,36 .
Despite the inherent difficulties of establishing the diet of extinct vertebrates with no suitable modern analogues such as pterosaurs, the following feeding habits have been proposed for derived pterodactyloids 37 , mainly based on their dentition (or absence of teeth) and the shape of their rostra: piscivory for Anhangueridae and their kin 30,[38][39][40] , Ikrandraco 41 , Pteranodontidae 30 , Nyctosauridae 42 , Chaoyangopteridae 43 , and Thalassodrominae 44 ; frugivory for Tapejarinae 45,46 ; durophagy for Dsungaripterus and related species 30 ; insectivory for Nemicolopterus 47 ; scavenging for Istiodactylus 21,22 ; cutting or ploughing through unconsolidated sediments for Argentinadraco 48 , and terrestrial stalking for the long-necked azhdarchids 49 . The dentition of Mimodactylus differs from all of them, suggesting that this lineage of derived pterodactyloid had a different feeding habit.
Studies on the shapes of teeth of extant insectivorous tetrapods emphasise that insectivorous species profit from having slimmer teeth that could be more easily used to breakdown arthropods due to the slight radius of curvature [50][51][52][53] . Within the pterosaur spectrum, the anurognathids that have well-spaced and isodont teeth have www.nature.com/scientificreports www.nature.com/scientificreports/ been regarded as insectivores 25,54 . Although wider, the tooth structure of Mimodactylus may suggest a similar feeding habit, allowing them to break up arthropod exoskeletons.
Aerial insectivory is closely linked to the ability to maneuver during flight [55][56][57] . Extant vertebrate aerial insectivores exhibit short wings with low aspect ratios that allow them to be highly maneuverable in the air [56][57][58] , contrary to Mimodactylus libanensis, which had long wings with high aspect ratio (Fig. 4b). In Mimodactylus, as open-sea flyers, the ability to maneuver during flight appears to be limited and it was likely high stable during flight as observed in albatrosses and other birds (Fig. 5). This might also have been the case for some large pterosaurs such as anhanguerians, istiodactylids and pteranodontians, which are considered to have conducted a dynamic soaring 58 . Therefore, alternately to the insectivore hypothesis, Mimodactylus and their relatives might also have been capable of foraging for decapod crustaceans on surface waters, just like some albatross species feed upon caridean or penaeid shrimps 59 . In addition, a broad rostrum 60 and spaced but relatively robust and pointed teeth 61,62 could be good tools to seize shrimps in the water.
Observing the fossil content of the Hjoûla Lagerstätte as well as of other Cretaceous Lebanese Lagerstätten, no insects were recovered so far 63,64 . Even terrestrial plants are extremely rare at Hjoûla 65 , suggesting that this Lebanese Lagerstätte was far away from emergent areas such as islands, with the continent several hundred kilometres away. On the other hand, decapod crustaceans are the most common invertebrates found in Hjoûla 66,67 . No taphonomic bias was detected to explain the absence of insects, pointing to fishes and zooplankton as potentially the main local source of food for pterosaurs. Mimodactylus libanensis also has a broad rostrum, which is consistent with a faunivorous feeding habit -or primarily feeding on crustaceans -as present in extant ducks, boat-billed herons, and shoebills 68 . Although insectivory cannot be ruled out, all available evidence suggests that Mimodactylus was feeding on crustaceans.
This new pterosaur lived in archipelagos and scattered islands, which were present during the Late Cretaceous in the gigantic carbonate platform bordering the northern part of the Afro-Arabian continent with the Neotethys (Fig. 5). The discovery of Mimodactylus libanensis expands the spectrum of possible feeding strategies in derived pterodactyloids, a group of fascinating volant reptiles for which we still know very little. phylogenetic analysis. We performed a phylogenetic analysis using the software TNT 1.5 69 . This analysis is based essentially on Holgado et al. 34 (for further details see Supplementary Information). Search for the most parsimonious trees (MPTs) was conducted via Traditional Search (TBR swapping algorithm), 10,000 replicates, random seed and collapsing trees after search.